“Anatomically-correct” dosimetric parameters may be better predictors for esophageal toxicity than are traditional CT-based metrics

Purpose: Incidental esophageal irradiation during lung cancer therapy often causes morbidity. There is interest in trying to relate esophageal dosimetric parameters to the risk of injury. These parameters typically rely on CT-defined esophageal contours, and thus systematic limitations in esophageal contouring will influence these parameters. We herein assess the ability of a correction method, based on physiologic principles, to improve the predictive power of dosimetric parameters for radiation-induced esophageal injury. Methods and Materials: Esophageal contours for 236 patients treated for lung cancer were quantitatively analyzed. All patients received three-dimensional planning, and all contours were generated by the same physician on axial CT images. Traditional dose-volume histogram (DVH)–based dosimetric parameters were extracted from the three-dimensional data set. A second set of “anatomically correct” dosimetric parameters was derived by adjusting the contours to reflect the known shape of the esophagus. Each patient was scored for acute and late toxicity using ROTG criteria. Univariate analysis was used to assess the predictive power of corrected and uncorrected dosimetric parameters (e.g., mean dose, V 50 , and V 60 ) for toxicity. The p values were taken as a measure of their significance. Results: The univariate results indicate that both corrected and uncorrected dosimetric parameters are generally predictors for toxicity. The corrected parameters are more highly correlated (lower p value) with outcomes than the uncorrected metrics. Conclusions: The inclusion of corrections, based on anatomic realities, to DVH-based dosimetric parameters may provide dosimetric parameters that are better correlated with clinical outcomes than are traditional DVH-based metrics.